Â鶹ÊÓƵ

Supervisors:

• Dr Himanshu Kaul  himanshu.kaul@leicester.ac.uk
•  Dr Kees Straatman
• Prof Yuhua Li (University of Cardiff)

Project Description:

Organoids are transforming research into understanding emergent biological patterns that shape development, regeneration, and disease. Their three-dimensional nature, physiological proximity to human (patho)biology, high-throughputness, and relative simplicity have made them highly desirable tools to capturing emergent biological patterns reproducibly and reliably1. More specifically, organoids offer us a robust platform for gathering information in great spatial and temporal resolution across multiple biological levels (protein, cell, organoid, organism), which in turn allows users to explore models, processes, and mechanisms in an integrated manner in a way that was not possible previously. Recent advances in live-cell microscopy is enabling researchers to monitor organoid growth, differentiation, cellular interactions, and responses to perturbations in real time. However, continuous live imaging generates large, complex, and often noisy datasets that are challenging to process, analyse, and interpret. The complex growth patterns of organoids when being imaged in a high-throughput manner makes data stitching and analysis at best labour intensive, and subject to bias, and at worst uninterpretable. However, recent developments in artificial intelligence (AI) and generative AI offer an unprecedented opportunity to transform the microscopy pipeline.

This PhD project aims to develop and validate an AI-enabled microscopy framework for live organoid imaging. The project will integrate a novel lung organoid protocol2 with deep learning3 and generative AI4 methods to improve image quality, reduce imaging burden, automate quantitative analysis and in particular time stitching of organoid data during growth, and generate predictive representations of organoid development. Ultimately, we aim to establish a robust platform yielding biologically meaningful spatiotemporally-relevant insights from longitudinal organoid datasets while minimising experimental constraints. This will be achieved via the following specific aims:

Aim 1: Build generative AI-supported automated pipelines for organoid segmentation, cell tracking, lineage reconstruction, and phenotype quantification using deep learning and transformer-based architectures
Aim 2: Develop predictive generative models capable of forecasting organoid growth, differentiation trajectories, and treatment responses from longitudinal imaging datasets
Aim 3: Benchmark performance against existing microscopy workflows

Expected project outcomes include new AI methods for live microscopy, open-source software tools for organoid analysis, and mechanistic insights into organoid development. More broadly, the project will contribute to the emerging field of AI-assisted experimental biology by establishing generative AI as a practical tool for image acquisition, analysis, and prediction. The resulting framework has the potential to reduce experimental costs, improve reproducibility, and increase the information extracted from complex biological imaging datasets. 

Successful candidate will work with engineers, AI scientists, biomedical scientists, and industrial stakeholders. They will receive interdisciplinary training in advanced microscopy, organoid biology, machine learning, computer vision, generative AI, scientific programming (Python/PyTorch), cloud and high-performance computing, data management, and reproducible research practices. The project will provide highly transferrable expertise valued across academia and industry.

References
1. Kaul H, et al. (2023). Virtual cells in a virtual microenvironment recapitulate early development-like patterns in human pluripotent stem cell colonies. Stem Cell Reports, 18, 377-293.
2. Derjean M, et al. (2026). Primary cells-derived lung organoids to enable drug design and optimization. ERJ Open Research, 12, PS211.
3. Moen E, et al. (2019). Deep learning for cellular image analysis. Nature Methods, 16, 1233–1246.
4. Lafarge MW, et al. (2024). Generative AI for biological imaging: opportunities and challenges. Nature Reviews Bioengineering, 5, 30.